Mitigating the Effect of Obstacles in Narrowband Ultrasonic Localization Systems

Abstract

This paper develops a method for mitigating the negative effects of obstacles in narrowband, time division multiple access (TDMA), ultrasonic localization systems. The method builds upon the robust Bayesian classifier for ultrasonic localization (RoBCUL) algorithm which utilizes an iteratively reweighted least squares (IRLS) scheme. This algorithm has the advantage of low computational cost but loses performance in the presence of obstacles. The improved version of the RoBCUL algorithm presented in this paper uses a statistical test applied after each iteration of the regression, using a weighted residual vector calculated from the weight matrix and residual vector. The technique was tested using experimental data with its performance being quantified by its ability to correctly classify all the signals received during a single TDMA cycle. The extended version performed significantly better in all obstacle scenarios than the original, correctly classifying 100% of TDMA cycles in the scenarios with no obstacles, 97.6% with one obstacle, and 89.0% with two obstacles.